Marin Ivezic
A quantum company’s pitch deck usually puts the intellectual property on one slide: 42 granted patents, 60 applications pending, filings across seven jurisdictions. That count is the easiest number in the deck to produce, and it’s close to the least informative. Two other numbers do more work. A utility patent’s term generally runs 20 years from the earliest effective non-provisional filing date it claims priority to, and in Europe from the actual filing date. Neither clock starts at grant, and in the United States a provisional application does not start it either. And in most jurisdictions an application publishes 18 months from the earliest claimed filing date, granted or not, though in the United States an applicant who does not intend to file abroad can request nonpublication. Neither number appears on the slide, and both change what the portfolio is worth.
Two readers look at the same document for different reasons. An investor wants to know whether the position can be held long enough to matter commercially. A procurement team wants to know whether buying the product creates exposure of its own. The questions below serve both. You don’t need a patent attorney in the room to ask most of them, though a few can only be answered with one.
Why the count is the weakest signal
Filing volume responds to incentives at least as much as it responds to invention. Where a government subsidises applications, where a university ties promotion to filings, or where a national programme sets a numeric target, filings rise without a matching rise in defensible technology. That’s a statement about what a count measures, not about the quality of anyone’s research.
Three distortions do most of the damage in practice.
Families counted as documents. One invention filed in the United States, Europe, Japan, China and under the international system is one patent family and five or more documents. A portfolio described as 40 patents may be eight inventions with wide geographic coverage, which is a perfectly respectable position, or it may be 40 unrelated filings of uneven quality. The families are the unit of analysis. Ask for the list by family, with the priority date of each.
Applications counted as rights. A pending application confers no right to stop anyone. Many quantum applications never grant, and those that do frequently grant with claims much narrower than the ones filed. Until you’ve seen the granted claims, you’ve seen a wish.
Territory ignored. Patents are national. A grant in one country does nothing in another, and a company that filed only at home has no position in your market. This matters most for buyers: a supplier can hold a genuinely strong domestic portfolio and still be entirely unprotected in the country where you plan to deploy the system.
Counting across jurisdictions is the standard way a modest portfolio is made to look large in a deck, and it works because the question about families almost never gets asked.
Five questions that change the answer
What is claimed, an apparatus or a method? Quantum companies produce both hardware and algorithms, and the two protect very differently. Section 101 of the U.S. patent statute governs what kinds of things can be patented at all, and abstract ideas, including bare mathematics, fall outside it. A compilation technique or an error-mitigation routine described as a sequence of calculations invites that objection. The same technique claimed as a method carried out on specific control hardware usually survives it. So a software-led quantum company can hold a portfolio that is thinner than its actual advantage, while a hardware company’s claims read directly onto the machine it ships. Read the independent claims and ask whether they describe the product being sold or something adjacent to it.
Where is it granted, and is that where the money is? Map the grants against the revenue plan. A device manufacturer selling into Europe with grants only in the United States has a marketing asset rather than a commercial one. The gap is often deliberate and defensible, because national-phase filings are expensive and a small company has to choose. What you want to see is that somebody chose, and can say why.
How much term is left? Deep-tech portfolios frequently rest on a priority chain that started years before the product did. Continuation practice, where a later application claims the original priority date to pursue additional claims, keeps a family alive but does not extend the clock: the 20 years still run from that earliest effective filing date. A company whose core family has a 2016 priority date and whose product reaches general availability in 2027 has under a decade of exclusivity on a technology that may need most of that decade to find its market. A ten-year runway on a decade-long product cycle just means the investment case rests on the next generation of filings, so ask what is in the pipeline and who is writing it.
Is the real advantage in the patents or in the know-how? In quantum hardware, much of what separates a working device from a demonstration is process: fabrication recipes, materials handling, calibration routines, the accumulated corrections that lift yield. Very little of that is patented, because patenting means publishing it, and much of it cannot be reverse-engineered from a delivered system. Trade secrecy is often the right call. It also concentrates the value in a small number of people. For an investor, that converts an IP question into a retention question. For a buyer, it becomes a continuity question, and the awkward version is worth asking directly: if three named engineers leave, what happens to device yield, and what happens to our maintenance contract?
Does the product have to interoperate? This one determines whether a patent can ever be a revenue stream. Where customers buy a sealed box, an exclusive position on the internals is straightforward to hold. Where the technology has to work with everyone else’s, patents tend to be pulled into standards, and standards bodies attach licensing conditions. Commitments described as FRAND, meaning fair, reasonable and non-discriminatory, oblige the holder to license to all comers on terms that are not designed to exclude. A patent under a FRAND commitment is worth something. It isn’t worth what an exclusive position on the same technology would be worth.
When the standard eats the patent
Post-quantum cryptography is the clearest case available, and it is instructive for anyone underwriting a quantum-adjacent business case.
ML-KEM (previously Kyber), the key-encapsulation mechanism standardised by NIST, is meant to be implemented in essentially every device that establishes an encrypted connection. A technology with that reach cannot carry a royalty, because the royalty would simply prevent adoption. NIST’s competition therefore required licensing commitments from submitters, and before ML-KEM could be published, third-party patent claims over lattice-based key exchange had to be resolved so the standard could be implemented without a royalty obligation. The technical merit of the algorithm was never the constraint at that stage; the terms of use were.
The lesson generalises beyond cryptography. Any layer of the quantum stack that must interoperate, including control interfaces, networking protocols and cloud access APIs, is likely to travel the same road. Patents there become bargaining positions inside a standards process rather than exclusive rights in a market. Patents on the parts nobody else needs to interoperate with, a qubit fabrication method or a cryogenic packaging design, keep their teeth. When a business case rests on owning something that will have to become a standard, discount it.
Freedom to operate is the buyer’s problem too
Freedom to operate is the question of whether you can sell or use a product without infringing someone else’s rights. It’s separate from whether you hold patents of your own, and a company can own a strong portfolio and still infringe. In a field where many groups are converging on similar qubit designs and similar control electronics, that combination is common rather than exotic.
For procurement, the exposure travels with the purchase. An injunction against your supplier can strand a deployed system, and a damages claim can reach the user of an infringing device as well as its maker. Four things belong in the diligence pack.
- Whether a freedom-to-operate analysis exists for the product as configured for you, and when it was last refreshed. A search performed before a major redesign tells you little.
- The indemnity, and its cap. IP indemnities are frequently capped at contract value, which is the one figure guaranteed to be smaller than the loss if a system has to be pulled out.
- What the supplier has licensed in, and on what terms. A licence that terminates on change of control is a real risk in a sector where startups get acquired.
- The university licence chain, for any spin-out. Under the Bayh-Dole Act, U.S. universities may own inventions made with federal funding, and the government retains a nonexclusive licence plus march-in rights. Spin-outs then license that IP back, and the licence is usually limited by field of use, meaning the company may practise the invention in stated applications only. An exclusive licence for quantum sensing in medical imaging is not a licence to sell into defence. Investors should read the same document, along with its diligence milestones, since a missed milestone can convert exclusivity into something considerably less.
The thicket, and what all that filing also produces
A patent thicket is a dense mesh of overlapping rights that makes it hard to build anything without touching somebody’s claim. Superconducting qubit technology is the sub-field most often named as heading that way, and a startup working there does face claims held by several large companies and several universities at once.
There’s a counterweight, and it’s easy to miss. Most applications publish at 18 months, granted or not, and once published they create prior art that everyone can rely on. The main exception is the U.S. nonpublication request, available to applicants who do not plan to file abroad. Section 103, the obviousness test, then bars later patents on ideas that a skilled engineer would arrive at from what is already public. Thousands of quantum applications have published without ever becoming enforceable rights, and each one narrows what can still be claimed broadly. The practical effect is that broad claims filed late deserve scepticism, and a company asserting sweeping coverage over a well-populated technique probably has claims narrower than the assertion.
What a portfolio worth paying for looks like
The claims read onto the product the company actually sells, not onto an earlier prototype. The grants exist where the customers are. The remaining term overlaps the commercial window rather than expiring inside it. Ownership is clean, and any in-licence has been read to its field-of-use clause. Where the advantage is know-how instead of patents, somebody has written it down and thought about who holds it. That portfolio might be eight families. The one on the slide with 42 might be worth less.
Commercial judgment about quantum technology depends on this kind of reading, and it is a skill that transfers across procurement, investment and program governance. Quantum Academy’s current programs are listed at quantumacademy.com/. For deeper technical background on the patent picture itself, PostQuantum.com covers the underlying filing record in more detail.